R语言中使用spread()转置数据时为何出现NA缺失值？

我有如下数据：

dataA= data.frame(
  Location = rep(c("East", "West", "North"), each = 12 * 2),
  Season = rep(rep(c(2021, 2022), each = 6), 3),
  Genotype = rep(rep(c("CV1", "CV2"), each = 6), 3),
  Iron_ton_ha = rep(c(21.7127, 16.9369, 30.7798, 14.0833, 13.2023, 11.1327, 15.0424, 
                      9.6616, 5.8569, 6.1881, 8.0127, 6.0291), each = 6),
  Stage = rep(c("Vegetative", "Reproductive", "Maturity"), each = 12),
  Fe = rep(c(0.44, 0.36, 0.31, 0.28, 0.79, 0.38, 0.59, 0.35, 0.41, 0.4, 0.29, 0.27,
             0.37, 0.3, 0.31, 0.28, 0.16, 0.24, 0.57, 0.27, 0.23, 0.26, 0.26, 0.27,
             0.13, 0.15, 0.04, 0.06, 0.03, 0.23, 0.52, 0.06, 0.08, 0.11, 0.1, 0.04), 
           each = 2)
)

我希望将Stage列中的每个变量转为单独列，使用了以下代码，但结果出现大量NA缺失值，请问为何该代码无法正常工作？

使用的代码及结果：

library(dplyr)
library(tidyverse)

dataB= dataA %>%
  group_by(Location, Season, Genotype, Iron_ton_ha, Stage) %>%
  summarise(Fe = mean(Fe, na.rm = TRUE)) %>%
  spread(key = Stage, value = Fe)

dataB

# A tibble: 12 × 7
# Groups:   Location, Season, Genotype, Iron_ton_ha [12]
   Location Season Genotype Iron_ton_ha Maturity Reproductive Vegetative
   <chr>     <dbl> <chr>          <dbl>    <dbl>        <dbl>      <dbl>
 1 East       2021 CV1            21.7   NA            NA          0.37 
 2 East       2021 CV1            30.8   NA             0.45      NA    
 3 East       2022 CV2            14.1   NA             0.32      NA    
 4 East       2022 CV2            16.9   NA            NA          0.483
 5 North      2021 CV1             5.86  NA             0.107     NA    
 6 North      2021 CV1             8.01   0.22         NA         NA    
 7 North      2022 CV2             6.03   0.0833       NA         NA    
 8 North      2022 CV2             6.19  NA             0.107     NA    
 9 West       2021 CV1            13.2    0.327        NA         NA    
10 West       2021 CV1            15.0   NA            NA          0.357
11 West       2022 CV2             9.66  NA            NA          0.263
12 West       2022 CV2            11.1    0.227        NA         NA    

原因分析 #

核心问题出在group_by的参数里——你把Stage也加入了分组条件。这会导致每个分组只对应单个Stage值，后续用spread转置时，每个分组自然只有当前Stage的列有计算结果，其他Stage列只能填充NA。

比如East-2021-CV1-21.7-Vegetative这个分组只计算了Vegetative阶段的Fe均值，转置后Maturity和Reproductive列没有对应数据，只能显示NA。

正确代码 #

方法1：用pivot_wider（tidyverse推荐替代spread的函数） #

先按除Stage外的维度分组统计，再转置：

library(tidyverse)

dataB <- dataA %>%
  group_by(Location, Season, Genotype, Iron_ton_ha, Stage) %>%
  summarise(Fe = mean(Fe, na.rm = TRUE), .groups = "drop") %>% # 取消分组避免后续问题
  pivot_wider(names_from = Stage, values_from = Fe)

方法2：修正spread的用法 #

如果坚持用spread，必须先取消分组，否则spread会在每个子分组内单独转置，依然会出现NA：

library(dplyr)
library(tidyr)

dataB <- dataA %>%
  group_by(Location, Season, Genotype, Iron_ton_ha, Stage) %>%
  summarise(Fe = mean(Fe, na.rm = TRUE)) %>%
  ungroup() %>% # 关键步骤：取消分组
  spread(key = Stage, value = Fe)

结果说明 #

修正后，每个Location, Season, Genotype, Iron_ton_ha组合对应一行数据，三个Stage列（Maturity、Reproductive、Vegetative）都会显示对应阶段的Fe均值，不会再出现大量NA。

内容的提问来源于stack exchange，提问作者J.K Kim